[
https://issues.apache.org/jira/browse/MAPREDUCE-7435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17714187#comment-17714187
]
ASF GitHub Bot commented on MAPREDUCE-7435:
-------------------------------------------
steveloughran commented on PR #5519:
URL: https://github.com/apache/hadoop/pull/5519#issuecomment-1514982508
updated pr has been run through azure, with stats of a terasort being
```
2023-04-19 16:15:23,305 INFO [JUnit-test_140_teracomplete]:
statistics.IOStatisticsLogging
(IOStatisticsLogging.java:logIOStatisticsAtLevel(269)) - IOStatistics:
counters=((commit_file_rename=4)
(committer_bytes_committed=200021)
(committer_commit_job=3)
(committer_files_committed=4)
(committer_task_directory_depth=7)
(committer_task_file_count=8)
(committer_task_file_size=200021)
(committer_task_manifest_file_size=127578)
(job_stage_create_target_dirs=3)
(job_stage_load_manifests=3)
(job_stage_rename_files=3)
(job_stage_setup=3)
(op_create_directories=3)
(op_delete=3)
(op_get_file_status=13)
(op_get_file_status.failures=13)
(op_list_status=10)
(op_load_all_manifests=3)
(op_load_manifest=7)
(op_mkdirs=13)
(op_msync=3)
(task_stage_commit=7)
(task_stage_scan_directory=7)
(task_stage_setup=7));
gauges=((stage.job_stage_create_target_dirs.free.memory=1542076752)
(stage.job_stage_create_target_dirs.heap.memory=533055152)
(stage.job_stage_create_target_dirs.total.memory=2075131904)
(stage.job_stage_load_manifests.free.memory=1544775808)
(stage.job_stage_load_manifests.heap.memory=530356096)
(stage.job_stage_load_manifests.total.memory=2075131904)
(stage.job_stage_rename_files.free.memory=1505757416)
(stage.job_stage_rename_files.heap.memory=569374488)
(stage.job_stage_rename_files.total.memory=2075131904)
(stage.setup.free.memory=1688139168)
(stage.setup.heap.memory=386992736)
(stage.setup.total.memory=2075131904));
minimums=((commit_file_rename.min=53)
(committer_task_directory_count=0)
(committer_task_directory_depth=1)
(committer_task_file_count=0)
(committer_task_file_size=0)
(committer_task_manifest_file_size=18018)
(job_stage_create_target_dirs.min=3)
(job_stage_load_manifests.min=184)
(job_stage_rename_files.min=73)
(job_stage_setup.min=267)
(op_create_directories.min=0)
(op_delete.min=30)
(op_get_file_status.failures.min=24)
(op_list_status.min=170)
(op_load_all_manifests.min=73)
(op_load_manifest.min=54)
(op_mkdirs.min=26)
(op_msync.min=0)
(task_stage_commit.min=176)
(task_stage_scan_directory.min=176)
(task_stage_setup.min=52));
maximums=((commit_file_rename.max=62)
(committer_task_directory_count=0)
(committer_task_directory_depth=1)
(committer_task_file_count=1)
(committer_task_file_size=100000)
(committer_task_manifest_file_size=18389)
(job_stage_create_target_dirs.max=4)
(job_stage_load_manifests.max=250)
(job_stage_rename_files.max=74)
(job_stage_setup.max=295)
(op_create_directories.max=1)
(op_delete.max=42)
(op_get_file_status.failures.max=113)
(op_list_status.max=189)
(op_load_all_manifests.max=139)
(op_load_manifest.max=125)
(op_mkdirs.max=74)
(op_msync.max=0)
(task_stage_commit.max=194)
(task_stage_scan_directory.max=194)
(task_stage_setup.max=93));
means=((commit_file_rename.mean=(samples=4, sum=227, mean=56.7500))
(committer_task_directory_count=(samples=14, sum=0, mean=0.0000))
(committer_task_directory_depth=(samples=7, sum=7, mean=1.0000))
(committer_task_file_count=(samples=14, sum=8, mean=0.5714))
(committer_task_file_size=(samples=7, sum=200021, mean=28574.4286))
(committer_task_manifest_file_size=(samples=7, sum=127578, mean=18225.4286))
(job_stage_create_target_dirs.mean=(samples=3, sum=11, mean=3.6667))
(job_stage_load_manifests.mean=(samples=3, sum=638, mean=212.6667))
(job_stage_rename_files.mean=(samples=3, sum=220, mean=73.3333))
(job_stage_setup.mean=(samples=3, sum=845, mean=281.6667))
(op_create_directories.mean=(samples=3, sum=2, mean=0.6667))
(op_delete.mean=(samples=3, sum=107, mean=35.6667))
(op_get_file_status.failures.mean=(samples=13, sum=638, mean=49.0769))
(op_list_status.mean=(samples=10, sum=1431, mean=143.1000))
(op_load_all_manifests.mean=(samples=3, sum=287, mean=95.6667))
(op_load_manifest.mean=(samples=7, sum=546, mean=78.0000))
(op_mkdirs.mean=(samples=13, sum=580, mean=44.6154))
(op_msync.mean=(samples=3, sum=0, mean=0.0000))
(task_stage_commit.mean=(samples=7, sum=1279, mean=182.7143))
(task_stage_scan_directory.mean=(samples=7, sum=1279, mean=182.7143))
(task_stage_setup.mean=(samples=7, sum=505, mean=72.1429)));
```
once abfs adds iostats context update of input stream reads, we could
collect and add that into the stats too; not worrying about it until then.
> ManifestCommitter OOM on azure job
> ----------------------------------
>
> Key: MAPREDUCE-7435
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-7435
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: client
> Affects Versions: 3.3.5
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Major
> Labels: pull-request-available
>
> I've got some reports of spark jobs OOM if the manifest committer is used
> through abfs.
> either the manifests are using too much memory, or something is not working
> with azure stream memory use (or both).
> before proposing a solution, first step should be to write a test to load
> many, many manifests, each with lots of dirs and files to see what breaks.
> note: we did have OOM issues with the s3a committer, on teragen but those
> structures have to include every etag of every block, so the manifest size is
> O(blocks); the new committer is O(files + dirs).
> {code}
> java.lang.OutOfMemoryError: Java heap space
> at
> org.apache.hadoop.fs.azurebfs.services.AbfsInputStream.readOneBlock(AbfsInputStream.java:314)
> at
> org.apache.hadoop.fs.azurebfs.services.AbfsInputStream.read(AbfsInputStream.java:267)
> at java.io.DataInputStream.read(DataInputStream.java:149)
> at
> com.fasterxml.jackson.core.json.ByteSourceJsonBootstrapper.ensureLoaded(ByteSourceJsonBootstrapper.java:539)
> at
> com.fasterxml.jackson.core.json.ByteSourceJsonBootstrapper.detectEncoding(ByteSourceJsonBootstrapper.java:133)
> at
> com.fasterxml.jackson.core.json.ByteSourceJsonBootstrapper.constructParser(ByteSourceJsonBootstrapper.java:256)
> at com.fasterxml.jackson.core.JsonFactory._createParser(JsonFactory.java:1656)
> at com.fasterxml.jackson.core.JsonFactory.createParser(JsonFactory.java:1085)
> at
> com.fasterxml.jackson.databind.ObjectMapper.readValue(ObjectMapper.java:3585)
> at
> org.apache.hadoop.util.JsonSerialization.fromJsonStream(JsonSerialization.java:164)
> at org.apache.hadoop.util.JsonSerialization.load(JsonSerialization.java:279)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.files.TaskManifest.load(TaskManifest.java:361)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.impl.ManifestStoreOperationsThroughFileSystem.loadTaskManifest(ManifestStoreOperationsThroughFileSystem.java:133)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.AbstractJobOrTaskStage.lambda$loadManifest$6(AbstractJobOrTaskStage.java:493)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.AbstractJobOrTaskStage$$Lambda$231/1813048085.apply(Unknown
> Source)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding.invokeTrackingDuration(IOStatisticsBinding.java:543)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding.lambda$trackDurationOfOperation$5(IOStatisticsBinding.java:524)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding$$Lambda$217/489150849.apply(Unknown
> Source)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding.trackDuration(IOStatisticsBinding.java:445)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.AbstractJobOrTaskStage.loadManifest(AbstractJobOrTaskStage.java:492)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.LoadManifestsStage.fetchTaskManifest(LoadManifestsStage.java:170)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.LoadManifestsStage.processOneManifest(LoadManifestsStage.java:138)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.LoadManifestsStage$$Lambda$229/137752948.run(Unknown
> Source)
> at
> org.apache.hadoop.util.functional.TaskPool$Builder.lambda$runParallel$0(TaskPool.java:410)
> at
> org.apache.hadoop.util.functional.TaskPool$Builder$$Lambda$230/467893357.run(Unknown
> Source)
> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:750)
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
